Sophi, a suite of artificial intelligence tools developed by The Globe and Mail, Canada has won this year’s Online Journalism Award for Technical Innovation in the Service of Digital Journalism. The award is given by the Online News Association (ONA) – the world’s largest digital journalism association. In May, Sophi won the 2020 North American Digital Media Award given by The World Association of News Publishers (WAN-IFRA) in the category of Best Digital News Start-up.
It is an artificial intelligence system that works autonomously to find and promote the most valuable content. ‘Valuable’ in this context refers to articles that can drive subscriptions/retain subscribers – they may not always be the ones that have the highest pageviews.
Launching a 175-year-old newsroom into the digital era
The Globe is one of Canada’s leading newspapers with 6M readers. It was founded 175 years ago and its journey in developing Sophi is an inspiring story with several valuable insights for other publishers.
How do you take a 175-year-old newsroom and successfully launch it into the digital era? This is the challenge that the Globe and Mail faced in 2012, when we set up our paywall so that we could continue to fund ambitious, nation-building journalism.Sophi.io
The publisher put up a paywall in 2012. One of the first dilemmas it was confronted with was understanding the difference in value between an article that was behind a paywall and brought in dozens of subscribers versus one that had thousands of page-views.
There were other issues as well. The newsroom was fixated on page views as till then that had been the only measure of digital success, because of the advertising revenue associated with it. However, after paywall implementation subscription revenue became increasingly important.
“What is the most valuable content?”
It was difficult for the journalists to understand the value of an article that was behind the paywall since it did not bring many page views. Moreover, only a few people could take a deep dive into data because the publisher was using complex tools that were not user-friendly for journalists.
“Before Sophi, The Globe used all kinds of analytics tools which showed what the most popular content was,” writes Sonali Verma, Senior Product Manager, Analytics, The Globe and Mail. “But none of them really answered the question that Sophi answers, which is: What is the most valuable content?”
The publisher hired data scientists to develop an artificial intelligence platform for natural language processing, prediction and optimization that could be easily used by its journalists. Sophi was developed by the data scientists in close collaboration with the newsroom. They devised a method of scoring every piece of content according to how much value it brought to the business.
The left-hand side of the bar above reflects the different ways in which an article can bring in advertising revenue. The right-hand side shows the role of an article in driving new subscriptions as well as in retaining existing subscribers.
Sophi also solves the problem of promotion bias. For example, by looking at the data, the editors can figure out whether promoting an article on a certain section page is paying-off. Further, if an article is successful on search or social referrals, they can see which attributes they could replicate in order for the next such article to be successful.
“Understand what to do more of and what to do less of”
Sophi helps the newsroom see that there are many ways in which an article can be valuable. “Every reporter and editor understands whether their content typically helps drive subscriptions or helps retain existing subscribers or helps bring new readers to the Globe through various sources, such as search or social,” according to an article on Sophi.io.
“It helped assigning editors understand what to do more of and what to do less of, since promotion bias was eliminated (so, it is no longer the case where an article that received a great deal of promotion and is, therefore, vastly popular, is considered the sort of article that we decide to do more of).”
Sophi’s AI engine has been trained by The Globe’s editors to understand what is relevant to each page and which articles are most appropriate. It understands how much each story contributes to subscriber retention, subscriber acquisition, registration potential and advertising dollars. It looks beyond just page-views, thus avoiding undesirable click-bait.
“Journalists can focus on finding and telling great stories”
This allows it to work autonomously and now more than more than 99% of the content on The Globe’s digital pages is placed by Sophi. It looks at the stories published across the website every 10 minutes to select the ones that merit greater promotion and updates the website accordingly. It also determines what to promote on The Globe’s social media pages and in its newsletters.
According to Verma, they found in A/B testing against traditional curation that click-through rates off the home page were up more than 17%. This helped lift the subscriber acquisition rate by more than 10%.
The Globe has also turned print laydown over to Sophi. It accomplishes in a few seconds what two people used to take two hours to do, freeing up valuable newsroom talent.
The newsroom of the future is one where journalists can focus on finding and telling great stories – something that machines can’t do.David Walmsley, Editor-in-Chief, The Globe and Mail
“This is why we asked our data scientists to automate the web pages, slowly and carefully testing the results before gradually implementing it across practically the entire site. And I’m very happy with the results,‘’ adds Walmsley.
The publisher has been using Sophi for more than a year, and according to Verma “no reader has complained or asked if a computer was behind the site.” She adds that “The Globe is also generating more advertising revenue due to increased inventory availability.”
All this has encouraged The Globe and Mail to start licensing Sophi to other news organizations, in the process opening up a new revenue stream for the publisher.